We check out the Synthesis source first, then the SyncEvolution source.
That let's us use SyncEvolution's autogen.sh/gen-autotools.sh scripts
for checking whether both sources are clean and properly tagged.
For Synthesis, we have to set the SYNTHESISSRC env variable.
Previously, the valgrind command was inserted in front of "env",
which always succeeded because child processes were not checked
(would include EDS). Now we run exactly the right process under
valgrind.
Instead of fetching a specific repo, use just "git fetch" to
pull updates into an existing repo. Specifying the repo
should have worked, but apparently is not done correctly (not
yet?) by the git in Debian Stable.
Using "git fetch" assumes that the URL of the repo does not
change between updates. That should be the case for us.
In a setup where it was run inside schroot we got error
messages from schroot because gconfd got stuck and prevented
a clean unmount of /tmp. Even "kill -9" wouldn't get rid of
such a hanging gconfd.
As a workaround, running the setup command twice seems to
succeed, including getting rid of the handing gconfd.
Previously do cleanup operation after implementing
a syncevolution setup. But if this setup raises
exceptions, we might get incorrect test results.
so clean all these files before doing any other
preparations
Checking out a branch which only exists remotely requires
a bit more work. Checking out tags might also have failed
due to an unnecessary "git pull". Now "git tag" and "git branch"
are used to check what the requested revision is and do the
necessary opertions.
The "prebuilt" test ran, but because it used the build directory
of the current platform it didn't find any backends, or the wrong
ones. Fixed by only using the backend setting from the previous
testing if the directory really exists and falling back to the
relative "backends" directory otherwise.
It might be possible to always use "backends". In order to not
change the running testing this wasn't done.
Mobical has a known server issue that it initiates a
slow sync when one client syncs with it in one-way-from-client
mode from the second time.
Because Mobical has no schedule on this, we skip this
test for Mobical interoperability test.
This change was originally committed a while back to the
runtest.py used in the nightly testing. Now we are merging
it back to the public runtest.py.
This patch adds the "Resend" test to various _SKIP and
_FAILURES lists and adds CLIENT_TEST_RESEND_TIMEOUT and
CLIENT_TEST_INTERRUPT_AT.
The long and unreadable environment strings for client-test
are now in the Python code as multiple independent strings
which get concatenated by Python:
function("foo"
"bar")
checkinstall *does* work on a Debian system, even under fakeroot.
The trick is to make /usr/src/rpm/RPMS writable for normal users.
The Makefile then copies the resulting .rpm back to the current
working directory.
The "lpia" (Low Power Intel Architecture) .debs are our normal
x86 binaries packaged with a different architecture name. The check
whether runtests.py was building for x86 broke when moving to an
64 bit bost. It must be run inside the build environment.
1) extract test results from many result files and combine them
into a xml format file to summary each unit test status
2) use 'compare.xsl' to compare two days' result and generate
comparison result
3) generate more visiable html output file by nightly result
and comparison result to make tracking regressions easily
First, the test really shouldn't fail. Need to investigate why it does on
one test platform. Second, it was applied to all test runs, overwriting
the desired value of CLIENT_TEST_FAILURE for that run.
These files used to contain the complete sync session
output. Now they contain just the user-visible output,
so don't compress anymore because that doesn't save much
space and makes the files harder to read.
In this mode, Client::Sync is replaced with
just Client::Sync::<first source>::testItems.
This assumes that all servers support WBXML
with our test data.
Only contacts in vCard 2.1 are tested. Several
failures are ignored because Google does not support
them:
* refresh from client
* one-way from client
* XML format
These tests are run although we know that they
fail, just in case that at some point they start
to work.
vcard30::testCopy used to fail because adding new clients forced
unexpected slow syncs on all other clients. This should be fixed
now, so the FAILURES entry could be removed.
For weird reasons that I don't have the patience to analyze running
the client-test wrapper script as generated by libtool fails
to set up the search path for libsyncevolution.so. It works
for syncevolution.
As a workaround let's simply set LD_LIBRARY_PATH ourselves, regardless
whether it is needed.
These tests are both slow and tend to fail with these servers
because of deficiencies in the servers. Disabled these tests
as discussed in Bugzilla #4282.
distcheck is useful to ensure that all files are distributed,
the result builds and cleans up properly. Doing this with
building of the GUI is important to avoid surprises during
packaging.
This is currently broken because opening the default system
addressbook depends on e_cal_new_system_memos(), which is
not available in all Evolution versions.
It could be added, but first we should sort out issue #3732.